Derived Fields Preserve Fine-Scale Detail in Budgeted Neural Simulators

1. 1Identify the key fields to derive in the neural simulator
2. 2Implement a coarsen-quantize-decode pipeline with derived fields
3. 3Evaluate the trade-off between storage budget and simulation accuracy
4. 4Optimize the simulator architecture and training objectives to leverage derived fields

Researchers and engineers working on neural simulators, particularly those dealing with fluid dynamics and Navier-Stokes equations, can benefit from this approach to improve simulation accuracy under budget constraints

💡 Derived fields can reduce high-frequency error in neural simulators without increasing storage costs

💡 Derived fields can help preserve fine-scale detail in neural simulators under budget constraints